Fig illustrates the effect of

Fig. 5 illustrates the effect of the Péclet number on the computational results, showing the time-averaged velocity magnitude fields for the three-row bundle at Pe = 600 and 1200. It can be seen that flow structures predicted by the URANS method for these regimes differ both within the tube bundle and in the wake of the bundle. For the case of Pe = 600, the jets markedly change their direction during inline flow. The flow in the wake is asymmetrical, and the merging of the jets, in vasopressin receptor antagonist to the regime with the highest Péclet number, is unpaired: both a wide jet (via merging the three jets leaving the bundle) and a second one (a single one) form. The instantaneous velocity and temperature fields computed for the widely packed bundle at Pe = 1200 are shown in Fig. 6. The flow, compared to the one obtained for the three-row closely packed bundle, is significantly more chaotic; the degree of flow mixing is also noticeably higher here, due to more intensive cross-flows in the wide intertubular spaces. The thermal wake from the heated tubes diffuses rather quickly, and becomes shorter. These changes in the flow structure have, as shown below, a significant effect on the average characteristics of heat transfer. The computed values of the average Nusselt number Nu for an individual tube were obtained in the following way: where Tw is the average temperature of the heated tube surface, qw is the thermal heat flux density on the tube\'s surface, λ is the coefficient of thermal conductivity. Two methods of evaluating the average Nusselt number were used when processing the numerical data obtained for the bundles consisting of more than one tube row in the cross direction: in the first case, only the data for the central cylinder was used (the number is denoted by Nu, the same as in the experiments of Ref. [3]), while in the second case, the arithmetic mean of the Nusselt numbers obtained for all the heated cylinders located in the sixth row was computed (the mean value is denoted by ). In Fig. 7, the Nusselt number values computed for different bundle configurations are compared with the experimental data obtained for the flow of pure alkali metals in inline bundles with the package parameters under consideration [1]. Fig. 7a shows the Nusselt number values for the closely packed bundles in the configurations of three and five tube rows in cross direction, computed by two different methods. Additionally, the plot includes a generalized approximation Nu = Pe0.5, recommended by the authors of Refs. [3,4] for use in estimations of heat transfer for pure liquid metals flowing around inline and staggered tube bundles. Let us note here that the experiments with flowing molten lead and lead-bismuth eutectic (described in Refs. [5,6]), as well as the recent experiments with lead [13], have yielded the heat transfer coefficient values that are several times lower than those estimated by the dependence Nu = Pe0.5. The main reason for sleep movement discrepancy is that the oxide film on the surface of the construction material forms, together with the near-surface lead layer saturated by impurities and oxide of the coolant, a low-conductivity layer that prevents intense heat transfer (the so-called thermal contact resistance (TCR) appears). In case of pure alkali metals, the TCR effects were relatively small. The analysis of the computed data presented in Fig. 7a reveals that for the case of the single-row closely packed bundle the URANS method (actually the steady RANS solution) predicts for all the computed regimes a heat transfer intensity that is significantly lower compared with the experimental data [1]. The computed values are approaching the experimental ones with an increase in the number of rows. The best agreement is achieved at high values of the Péclet number. Furthermore, the influence of the number of rows and the method of computing the average heat flux is less pronounced here. Table 2 shows the Nusselt number values evaluated by two different methods from the computational results for the three-row closely packed bundle on computational grids with varying degrees of refinement. The differences in the results obtained for various grids do not exceed 5–7%, with the higher Nusselt values predicted for the two lower Péclet values on a coarser grid; on the contrary, the computations on a finer grid for Pe = 1000 and 1200 predict a higher intensity of average heat transfer.